Entity-conditioned sentence generation

Abstract

An example operation may include one or more of tuning a language model based on dependencies between an original data set and a paraphrase data set of the original data set, parsing and annotating the paraphrase dataset with entity identifiers of predefined entities to generate an annotated paraphrase dataset, additionally tuning the language model based on entity dependencies between the original data set and the paraphrase data set based on the annotated paraphrase dataset, and storing the additionally tuned language model in a storage device.

What is claimed is: 
     
         1 . An apparatus, comprising:
 one or more processors; and   one or more memory devices coupled to the one or more processors and configured to store a language model, wherein the one or more processors are configured to:
 tune the language model based on dependencies between an original data set and a paraphrase data set of the original data set, wherein the language model is a machine learning language model; 
 parse and annotate the paraphrase data set with entity identifiers of predefined entities to generate an annotated paraphrase data set; 
 iteratively execute the tuned language model on the annotated paraphrase data set to identify entity dependencies between sentences included in the original data set and entities annotated within the annotated paraphrase data set; 
 additionally tune the language model, by changing variables and weights of the language model, based on the entity dependencies between the original data set and the paraphrase data set based on the annotated paraphrase data set; 
 store the additionally tuned language model in the one or more memory devices; and 
 execute the additionally tuned model on a different data set to generate a plurality of additional paraphrase data sets based on entities in the different data set. 
   
     
     
         2 . The apparatus of  claim 1 ,
 wherein the one or more processors are further configured to:
 execute the language model on the original data set and the paraphrase data set to learn dependencies between sentences that are included in the original data set and paraphrases that are included in the paraphrase data set. 
   
     
     
         3 . The apparatus of  claim 1 ,
 wherein the one or more processors are further configured to:
 convert the original data set into a first vector and the paraphrase data set into a second vector and identify dependencies between vectorized sentences in the original data set and vectorized paraphrases in the paraphrase data set in vector space. 
   
     
     
         4 . The apparatus of  claim 1 ,
 wherein the one or more processors are further configured to:
 execute a natural language processing model on the paraphrase data set to annotate and parse the paraphrase data set. 
   
     
     
         5 . The apparatus of  claim 1 ,
 wherein the one or more processors are further configured to:
 parse and annotate the paraphrase data set based on crowdsourced feedback from a plurality of user devices. 
   
     
     
         6 . The apparatus of  claim 1 ,
 wherein the one or more processors are further configured to:
 execute a machine learning model on the different data set and the plurality of additional paraphrase data sets to train the machine learning model, and store the trained machine learning model in the one or more memory devices. 
   
     
     
         7 . The apparatus of  claim 1 ,
 wherein the language model is stored on a blockchain.   
     
     
         8 . A method, comprising:
 tuning a language model based on dependencies between an original data set and a paraphrase data set of the original data set;   parsing and annotating the paraphrase data set with entity identifiers of predefined entities to generate an annotated paraphrase data set;   iteratively executing the tuned language model on the annotated paraphrase data set to identify entity dependencies between sentences included in the original data set and entities annotated within the annotated paraphrase data set;   additionally tuning the language model, by changing variables and weights of the language model, based on the entity dependencies between the original data set and the paraphrase data set based on the annotated paraphrase data set;   storing the additionally tuned language model in a storage device; and   executing the additionally tuned model on a different data set to generate a plurality of additional paraphrase data sets based on entities in the different data set.   
     
     
         9 . The method of  claim 8 ,
 wherein the tuning comprises executing the language model on the original data set and the paraphrase data set to learn dependencies between sentences that are included in the original data set and paraphrases that are included in the paraphrase data set.   
     
     
         10 . The method of  claim 8 ,
 wherein the tuning comprises converting the original data set into a first vector and the paraphrase data set into a second vector and identifying dependencies between vectorized sentences in the original data set and vectorized paraphrases in the paraphrase data set in vector space.   
     
     
         11 . The method of  claim 8 ,
 wherein the parsing and the annotating is performed via execution of a natural language processing model.   
     
     
         12 . The method of  claim 8 ,
 wherein the parsing and annotating is performed based on crowdsourced feedback from a plurality of user devices.   
     
     
         13 . The method of  claim 8 ,
 wherein the method comprises executing a machine learning model on the different data set and the plurality of additional paraphrase data sets to train the machine learning model, and store the trained machine learning model in the storage device.   
     
     
         14 . The method of  claim 8 ,
 wherein the language model is stored on a blockchain.   
     
     
         15 . A non-transitory computer-readable medium storing a set of instructions, the set of instructions comprising:
 one or more instructions that, when executed by one or more processors of a device, cause the device to:   tune a language model based on dependencies between an original data set and a paraphrase data set of the original data set;   parse and annotate the paraphrase data set with entity identifiers of predefined entities to generate an annotated paraphrase data set;   iteratively execute the tuned language model on the annotated paraphrase data set to identify entity dependencies between sentences included in the original data set and entities annotated within the annotated paraphrase data set;   additionally tune the language model, by changing variables and weights of the language model, based on the entity dependencies between the original data set and the paraphrase data set based on the annotated paraphrase data set;   store the additionally tuned language model in a storage device; and   execute the additionally tuned model on a different data set to generate a plurality of additional paraphrase data sets based on entities in the different data set.   
     
     
         16 . The non-transitory computer-readable medium of  claim 15 ,
 wherein the one or more instructions, to cause the device to tune the language model, cause the device to:   execute the language model on the original data set and the paraphrase data set to learn dependencies between sentences that are included in the original data set and paraphrases that are included in the paraphrase data set.   
     
     
         17 . The non-transitory computer-readable medium of  claim 15 ,
 wherein the one or more instructions, to cause the device to tune the language model, cause the device to:   convert the original data set into a first vector and the paraphrase data set into a second vector and identifying dependencies between vectorized sentences in the original data set and vectorized paraphrases in the paraphrase data set in vector space.   
     
     
         18 . The non-transitory computer-readable medium of  claim 15 ,
 wherein the one or more instructions further cause the device to:
 execute a natural language processing model on the paraphrase data set to annotate and parse the paraphrase data set. 
   
     
     
         19 . The non-transitory computer-readable medium of  claim 15 ,
 wherein the one or more instructions further cause the device to:   parse and annotate the paraphrase data set based on crowdsourced feedback from a plurality of user devices.   
     
     
         20 . The non-transitory computer-readable medium of  claim 15 ,
 wherein the language model is stored on a blockchain.